48
Fast Imaging Trajectories: EPI and PROPELLER M229 Advanced Topics in MRI Holden H. Wu, Ph.D. 2019.04.25 Department of Radiological Sciences David Geffen School of Medicine at UCLA
Fast Imaging Trajectories:EPI and PROPELLER
M229 Advanced Topics in MRI Holden H. Wu, Ph.D.
2019.04.25
Department of Radiological Sciences David Geffen School of Medicine at UCLA
Class Business
• Homework 1 due 4/26 Fri by 5 pm
• Homework 2 due 5/3 Fri by 5 pm
• TA office hours - Zhaohuan: Thu 4/25, 3-5 pm, MP300 B119
• Project proposal due 5/10 Fri by 5 pm - 1 page template on website
• Final presentation on 6/13 Thu
Outline
• EPI1
• PROPELLER2
- Pulse sequence and design considerations
- Alternatives
- Artifacts and corrections
- Applications
1Mansfield P, J Phys C: Solid State Phys., 1977 2Pipe, JG, Magn. Reson. Med., 1999
Gradient Echoθ
Gy
RF
Gx
ADC
…
…
…
…T2* decay
TR
TE
• Utilization of transverse magnetization
- With Ts = 8 µs and Nx = 128, Tacq = 1.024 ms
- <2% of T2* in brain at 3 T!1
• Scan time - TGRE = Npe x TR - TR = 10 ms, Npe = 256:
TGRE = 2.56 sec
1Peters, et al., Proc ISMRM 2006
Multi-echo Gradient Echo
θ
Gy
RF
Gx
TE1
…
…
…
…
Can perform T2* mapping
ΔTE can be non-uniform
ADC T2* decay
TE2 TE3 TE4
TR
Gradient-Echo EPI
θ
Gy
RF
Gx
…
…
…
…T2* decayADC
“bipolar” readout
TR
TEeff
Design Basics
Readout Gradient Duration
# readout points (Nx)
# phase-encoding lines (Ny)
Readout bandwidth (BW)
Field of view (FOVx)
Maximum gradient strength (Gmax)
Maximum gradient slew rate (Smax)
• What species are you imaging?
- T2, T2*?
- Utilize transverse magnetization efficiently by sampling up to, e.g., 2 × T2* (100 ms) → Readout gradient duration in EPI
- Total readout durations of up to 100 ms
EPI Sequence Parameters
θ
Gy
RF
Gx
TEeff
…
…
…
…ADCTR
x Nshot
Echo train length (ETL)Echo spacing (ESP)
Number of shots (Nshot)Effective TE (TEeff)
ESP ETL
EPI Sequence ParametersReadout Gradient Design
ky
kx
Gx
(γ/2π)·Gread·Ts = Δkx ≤ 1/FOVx
Tread = Ts·Nx
Gread ≤ Gmax
ESP ≥ Tread + 2·Tramp
GreadTs
TreadTramp
no ramp sampling
prewinder
readout
Tramp = Gread/SRSR ≤ Smax
EPI Sequence ParametersReadout Gradient Design Example:
Gx
GreadTs
TreadTramp
no ramp sampling
Ts = 8 µs; Nx = 128; FOVx = 22 cm; SR = 120 T/m/s
Gread = 13.3 mT/mESP = 1.246 ms
Bernstein et al., Handbook of MRI Pulse Sequences, Ch 16.1
If Ts = 4 µsESP = 0.955 ms
If Ts = 8 µs and SR = 20 T/m/s
ESP = 2.354 ms
If Ts = 4 µs and SR = 20 T/m/s
ESP = 3.172 ms
(γ/2π)·Gread·Ts = 1/FOVx
ESP = (Ts·Nx) + 2·(Gread/SR)
EPI Sequence ParametersPhase Encoding Gradient Design
Phase Encoding Bandwidth
ky
kx
(γ/2π)·Area(Gblip, Tblip) = Δky ≤ 1/FOVy
PEbw = 1/ESP ~ 1 kHz; more off-resonance artifacts
cf. RObw up to 500 kHz (Ts = 2 µs)
Gx
Gy
Tblip
Gblip
prewinder
blip
no ramp sampling
EPI Sequence Parameters
• ETL can be 4-64 or higher - Limited by T2* decay, off-resonance effects - aka “EPI factor”
• ESP typically ~1 ms - Must accommodate gradients and ADC - Short ESP facilitates high ETL
• Example: readout until S = 0.2 S0 - S = S0 * exp(-t/T2*); assume T2* = 60 ms - t = 96.6 ms - ESP = 1 ms; ETL = 96
Minimizing Readout Duration / ESP
• Higher gradient amplitudes and slew rates
• Higher readout bandwidths
• Sampling along the ramps
• Partial k-space acquisition - in x: “partial Fourier” < 1 - in y: phase FOV can be < 1
• Parallel imaging
• Inner volume imaging
Gradient-Echo EPI
SS
PE
RO
Gy
RF
…
…
…Gxp
Gx
Gyp
Gz
…
90oTEeff
FID
SE EPI
SS
PE
RO
Gz2
Gzc
Gy
180o
RF
…
…
…
…
Gxp Gx
Gyp
Gz
90oTEeff
Spin-Echo EPI
Spin Echo
Comparison
GRE-EPI SE-EPI
• GRE-EPI More signal dropouts, distortion
• GRE-EPI: More susceptibility effects, better for functional MRI acquisition
Managing EPI distortion
SE-EPI ???
Multi-shot EPI
• Single-shot EPI (ssEPI) - minimal motion artifacts - low resolution - geometric distortion and signal loss
• Multi-shot EPI (msEPI) - aka interleaved or segmented EPI - higher resolution - less distortion & signal loss (improve PEbw)- motion and phase inconsistencies
Multi-shot EPI
Bernstein et al., Handbook of MRI Pulse Sequences, Ch 16.1
ky
kx
Echo Time Shifting:
Comparison
ssEPI msEPI
Multi-shot EPI
(a) (b)
Interleaved Readout Segmented
Courtesy of Dr. Novena Rangwala
EPI Scan Time
• Scan time - Recall TGRE = Npe x TRGRE - Nshot = Npe / ETL - TEPI = Nshot x TREPI = (TGRE / ETL) x (TREPI/TRGRE)
• Example 1 - Npe = 256; ETL = 16; Nshot = 16 - TR = 30 ms: TEPI = 480 ms
• Example 2 - Npe = 64; ETL = 64; Nshot = 1 - TR = 100 ms: TEPI = 100 ms
Fly-Back GRE-EPI
θ
Gy
RF
Gx
…
…
…
…ADCTR
TEeff “unipolar” readout
Fly-Back GRE-EPI
• “Fly-back” gradients - No data sampling - Use max gradient amplitude/slew rate
• Advantages - All readouts in the same direction, minimal
artifacts
• Disadvantages - Longer ESP than bipolar EPI
Related Sequences
• 3D echo-volume imaging (EVI)
• Hybrid EPI + non-Cartesian (e.g., PROPELLER, EPI in a circular plane)
• Multi-echo chemical shift imaging
• Echo-planar spectroscopic imaging (EPSI), 2D and 3D
EPI Artifacts
• Nyquist ghosting artifacts
• Chemical-shift artifacts, e.g., fat
• Signal drop-out
• Geometric distortion
‘Orthogonal’ Plane ‘Oblique’ Plane
Source: Zhou, et al., ISMRM Abstracts, 1996, p. 386 Zhou, et al., 1997, US Patent 5,672,969
Linear Phase Nyquist Ghost
Constant Phase Nyquist Ghost
Oblique Nyquist Ghost
EPI Ghosting Artifacts
EPI Ghosting Artifacts
• Inconsistencies between even/odd echoes due to: - Spatially independent (constant):
B0 eddy currents, off-center freq mismatch - Linear and oblique phase errors:
k-space shifts from gradient / timing errors - Higher order eddy current effects - Concomitant magnetic fields
EPI Chemical Shift Artifacts
• Along readout - Δxcs = Δfcs·(FOVx / RObw) - At 1.5 T, ΔfWF ~ 210 Hz
for FOVx = 32 cm and RObw = 250 kHz,Δxcs = 0.027 cm
• Along phase encode - Δycs = Δfcs·(FOVy / ROpe),
ROpe = Nshot / ESP - for ESP = 1 ms, Δycs = 6.72 cm
EPI Considerations
• Minimize ESP (covered earlier)
• Spatial-spectral excitation for fat signal suppression
• Reconstruction steps - Row flipping and phase correction - Ramp sampling correction - Fourier transformation - (Possible) B0 inhomogeneity correction - (Possible) Gradient trajectory corrections
EPI Considerations
Addy NO et al., MRM 2012
Axial EPI, before & after trajectory correction
EPI Considerations
Image distortion and signal loss from dentures
Bernstein et al., Handbook of MRI Pulse Sequences, Ch 16.1
w/ dentures
Summary• Strengths
- very fast
• Challenges - T2* decay - high demand on slew rate - artifacts
Clinical Applications
• BOLD fMRI
• ASL
• DWI (see figure)
• Real-time MRI
• MRSI
• and more ...
b"="0"s/mm2" b"="750"s/mm2,"S/I"
b"="750"s/mm2,"R/L" b"="750"s/mm2,"A/P"
PROPELLER
• Periodically Rotated Overlapping ParallEL Lines with Enhanced Reconstruction1, aka BLADE
• Radial and Cartesian hybrid
• Oversampling at the center of k-space - correct inconsistencies between strips - reject data with through-plane motion - weigh strip contributions w.r.t. motion - average to decrease motion artifacts
1Pipe, MRM 1999; 42: 963-969
ky
kx
ky
kx
PROPELLER
2D Radial2D Cartesian
ky
kx
2D PROPELLER
always sampled
ky
kx
PROPELLERTrajectory Design:
N strips, successively rotated by dα = π/N
L lines per strip, M points per line
M
L
N
For an M x M image, need L·N = M·(π/2)central oversampled circle of diameter L
Scan time trade-offs based on L and N
Asymmetric FOV also possible
ky
kx
PROPELLERTrajectory Design Example:
24-cm FOV; 0.5 mm in-plan resln; L = 28
M = FOV/resln = 480
M
L
N
N = (M/L)·(π/2) ~ 27
TR = 4000 ms, Tscan = N·TR = 1 min 48 s
Bernstein et al., Handbook of MRI Pulse Sequences, Ch 17.5
ky
kx
PROPELLERTrajectory Design:
(c)
(d)
(a)
(b)
TSE
GRASE
Long axis EPI
Short axis EPI
PROPELLERTSE GRASE Long axis EPI
PROPELLERMotion correction:
Rotation in image space ⟷ rotation in k-space
Translation in image space ⟷ linear phase in k-space
Compare k-space magnitude between strips
Compare k-space phase between strips
Other motion in image space ⟷ k-space mag/phaseCompare and weigh importance of strips
PROPELLERReconstruction:
Pipe, MRM 1999; 42: 963-969
For each strip:
density compensation
PROPELLER
Pipe, MRM 1999; 42: 963-969
TSE, no motion TSE, motion
PROP, no motion PROP, motion PROP, motion, corr
PROP, corr k-space
T2 TSE BLADE T2 TSE
PROPELLER
Lavdas E, et al., MRI 2012; 30: 1099-1110
PROPELLER
• Advantages - robust to motion
• Disadvantages - increased scan time
• Extensions - 3D blocks; 3D rods (TORQ)
Clinical Applications
• Brain
• Abdomen/Pelvis
• MSK
• Diffusion-weighted imaging (high-resolution)
Summary
• EPI - very popular for fast MRI! - design, recon, corr drives a lot of research
• PROPELLER - very robust to motion - philosophy can be adapted to other seq
• Next time: Non-Cartesian sampling
Thanks!
• Further reading - Bernstein et al., Handbook of MRI Sequences - pubmed.org
• Acknowledgments - Novena Rangwala
Holden H. Wu, Ph.D.
http://mrrl.ucla.edu/wulab
